Wind Speed Forecasting Of Artificial Intelligent Optimizing Parameter Research And Application

Wind speed is key information to help market participants or managers involved in electricity market to develop energy policy and decision-making. However, because of high volatility, non-stationary, nonlinearity of series and the difficulty of obtaining enough information in time, the wind speed forecasting is a great challenge. Therefore, how to make an accurate prediction within limited information becomes a troublesome problem. In this paper, firstly, a combination approach is proposed, involving pre-disposed method (53Hanning Filter and wavelet denoising) to get rid of outliers and to denoise. And then by using the pre-disposed data,3-step wind speed forecasting is made on the basis of three common artificial intelligent statistical techniques, including Artificial Neural Network (BP and Elman), Support Vector Regression (SVR).Second, according to the result of first step that the pre-disposed SVR obtains the best accuracy among these models, in order to get a more accurate result, a new meta-heuristic algorithms (Cuckoo Search) is introduced to optimize the hyper-parameters of SVR. Third, a combination forecasting method based on the Cuckoo search is applied into this wind speed prediction so as to improve accuracy. Through data simulation, it is revealed that the proposed method in wind speed and electricity price forecasting has an improvement of accuracy.